AWS BEST PRACTICES FOR DDOS RESILIENCY

Page created by Kimberly Gross
 
CONTINUE READING
AWS BEST PRACTICES FOR DDOS RESILIENCY
AWS Best Practices for
  DDoS Resiliency

         June 2018
AWS BEST PRACTICES FOR DDOS RESILIENCY
© 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.

Notices
This document is provided for informational purposes only. It represents AWS’s current
product offerings and practices as of the date of issue of this document, which are subject to
change without notice. Customers are responsible for making their own independent
assessment of the information in this document and any use of AWS’s products or services,
each of which is provided “as is” without warranty of any kind, whether express or implied.
This document does not create any warranties, representations, contractual commitments,
conditions or assurances from AWS, its affiliates, suppliers or licensors. The responsibilities
and liabilities of AWS to its customers are controlled by AWS agreements, and this document
is not part of, nor does it modify, any agreement between AWS and its customers.
AWS BEST PRACTICES FOR DDOS RESILIENCY
Contents
Introduction                                           1
Denial of Service Attacks                              2
  Infrastructure Layer Attacks                         3
  Application Layer Attacks                            5
Mitigation Techniques                                  7
  Infrastructure Layer Defense (BP1, BP3, BP6, BP7)   10
  Application Layer Defense (BP1, BP2, BP6)           14
Attack Surface Reduction                              16
  Obfuscating AWS Resources (BP1, BP4, BP5)           16
Operational Techniques                                19
  Visibility                                          19
  Support                                             22
Conclusion                                            24
Contributors                                          24
Document Revisions                                    24
Appendix A: Additional Resources                      26
Abstract

This paper is intended for customers who want to improve the resiliency of their applications
running on Amazon Web Services (AWS) against Distributed Denial of Service (DDoS) attacks. It
provides an overview of DDoS attacks, capabilities provided by AWS, mitigation techniques,
and a DDoS-resilient reference architecture that can be used as a guide to help protect
application availability.

The paper is intended for IT decision makers and security engineers who are familiar with the
basic concepts of networking, security, and AWS. Each section has links to AWS
documentation that provides more detail on the best practice or capability.
Introduction

You work to protect your business from the impact of Distributed Denial of Service (DDoS)
attacks, as well as other cyberattacks. You want to keep your customers’ trust in your service
by maintaining the availability and responsiveness of your application. And you want to avoid
unnecessary direct costs when your infrastructure must scale in response to an attack.

AWS is committed to providing you with tools, best practices, and services to help ensure high
availability, security, and resiliency to defend against bad actors on the internet.

In this whitepaper, we provide you with prescriptive DDoS guidance. We describe different
attack types, such as infrastructure layer attacks and application layer attacks, and explain
which best practices are most effective to manage each attack type. We also outline the
services and features that fit into a DDoS mitigation strategy, and how each one can be used to
help protect your applications.

Page 1
Amazon Web Services – AWS Best Practices for DDoS Resiliency

Denial of Service Attacks

A Denial of Service (DoS) attack is a deliberate attempt to make your website or application
unavailable to users, such as by flooding it with network traffic. To achieve this, attackers use a
variety of techniques that consume large amounts of network bandwidth or tie up other
system resources, disrupting access for legitimate users. In its simplest form, a lone attacker
uses a single source to execute a DoS attack against a target, as shown in the following
diagram (Figure 1).

                                  Figure 1: Diagram of DOS attack

But in a Distributed Denial of Service (DDoS) attack, an attacker uses multiple sources – which
may be distributed groups of malware infected computers, routers, IoT devices, and other
endpoints– to orchestrate an attack against a target. As illustrated in the following DDoS
attack diagram (Figure 2), a network of compromised hosts participates in the attack,
generating a flood of packets or requests to overwhelm the target.

                                 Figure 2: Diagram of DDoS attack

DDoS attacks are most common at layers 3, 4, 6, and 7 of the Open Systems Interconnection
(OSI) model, which is described in following table (Table 1). Layer 3 and 4 attacks correspond
to the Network and Transport layers of the OSI model. We’ll refer to these collectively as
infrastructure layer attacks. Layer 6 and 7 attacks correspond to the Presentation and

Page 2
Amazon Web Services – AWS Best Practices for DDoS Resiliency

Application layers of the OSI model. We’ll address these together as application layer attacks.
We describe examples of these attack types in the next sections.

   #       Layer          Unit              Description               Vector Examples

   7   Application     Data         Network process to application   HTTP floods, DNS
                                                                     query floods
   6   Presentation    Data         Data representation and          TLS abuse
                                    encryption
   5   Session         Data         Interhost communication          N/A

   4   Transport       Segments     End-to-end connections and       SYN floods
                                    reliability
   3   Network         Packets      Path determination and logical   UDP reflection
                                    addressing                       attacks
   2   Data Link       Frames       Physical addressing              N/A

   1   Physical        Bits         Media, signal, and binary        N/A
                                    transmission

                        Table 1: Open Systems Interconnection (OSI) Model.

Infrastructure Layer Attacks

The most common DDoS attacks, UDP reflection attacks and SYN floods, are infrastructure
layer attacks. An attacker can use either of these methods to generate large volumes of traffic
that can inundate the capacity of a network or tie up resources on systems like a server,
firewall, IPS, or load balancer. While these attacks can be easy to identify, to effectively
mitigate them, you must have a network or systems that scale up capacity more rapidly than
the inbound traffic flood. This extra capacity is to either filter out or absorb the attack traffic
enabling your system and application to respond to your legitimate customer traffic.

UDP Reflection Attacks

UDP reflection attacks exploit the fact that UDP is a stateless protocol. Attackers can craft a
valid UDP request packet listing the attack target’s IP as the UDP source IP address. The
attacker has now falsified, or “spoofed”, the UDP request packet’s source IP. An attacker then
sends the UDP packet containing the spoofed source IP to an intermediate server. The server is
tricked into sending its UDP response packets to the targeted victim IP rather than back to the

Page 3
Amazon Web Services – AWS Best Practices for DDoS Resiliency

attacker’s IP address. The intermediate server is used because it generates a response that is
several times larger than the request packet effectively amplifying the amount of attack traffic
sent to the target IP address.

The amplification factor, which is the ratio of response size to request size, varies depending
on which protocol the attacker uses: DNS, NTP, or SSDP. For example, the amplification factor
for DNS can be 28 to 54 times the original number of bytes. So if an attacker sends a request
payload of 64 bytes to a DNS server, they can generate over 3400 bytes of unwanted traffic to
an attack target. The reflection tactic and amplification effect are illustrated in the following
diagram (Figure 3).

                                  Figure 3: UDP reflection attack

SYN Flood Attacks

When a user connects to a TCP service, like a web server, their client sends a SYN
(synchronization) packet. The server returns a SYN-ACK packet in acknowledgement, and,
finally, the client responds with an ACK packet, which completes the expected three-way
handshake. This typical handshake is illustrated in the following diagram (Figure 4):

Page 4
Amazon Web Services – AWS Best Practices for DDoS Resiliency

                                  Figure 4: SYN 3-way handshake

In a SYN flood attack, a malicious client sends a large number of SYN packets, but never sends
the final ACK packets to complete the handshakes. The server is left waiting for a response to
the half-open TCP connections and eventually runs out of capacity to accept new TCP
connections. This can prevent new users from connecting to the server. SYN floods can reach
up to 100s of Gbps, but the attack is not about SYN traffic volume but rather tying up available
server connections resulting in no resources for legitimate connections.

Application Layer Attacks

An attacker may target the application itself by using a layer 7 or application layer attack. In
these attacks, similar to SYN flood infrastructure attacks, the attacker attempts to overload
specific functions of an application to make the application unavailable or extremely
unresponsive to legitimate users. Sometimes this can be achieved with very low request
volumes that generate only a small volume of network traffic. This can make the attack
difficult to detect and mitigate. Examples of application layer attacks include HTTP floods,
cache-busting attacks, and WordPress XML-RPC floods.

In an HTTP flood attack, an attacker sends HTTP requests that appear to be from a real user of
the web application. Some HTTP floods target a specific resource, while more complex HTTP
floods attempt to emulate human interaction with the application. This can increase the
difficulty of using common mitigation techniques like request rate limiting.

Cache-busting attacks are a type of HTTP flood that uses variations in the query string to
circumvent content delivery network (CDN) caching. Instead of being able to return cached
results, the CDN must contact the origin server for every page request, and these origin
fetches cause additional strain on the application web server.

Page 5
Amazon Web Services – AWS Best Practices for DDoS Resiliency

With a WordPress XML-RPC flood attack, also known as a WordPress pingback flood, an
attacker misuses the XML-RPC API function of a website hosted on the WordPress content
management software to generate a flood of HTTP requests. The pingback feature allows a
web site hosted on WordPress (Site A) to notify a different WordPress site (Site B) that Site A
has created a link to Site B. Site B then attempts to fetch Site A to verify the existence of the
link. In a pingback flood, the attacker misuses this capability to cause Site B to attack Site A.
This type of attack has a clear signature: “WordPress” is typically present in the “User-Agent”
of the HTTP request header.

There are also other forms of malicious traffic that can impact an application’s availability.
Scraper bots automate attempts to access a web application to steal content or record
competitive information, like pricing. Brute force and credential stuffing attacks are
programmed efforts to gain unauthorized access to secure areas of an application. These are
not strictly DDoS attacks; but their automated nature can look similar to a DDoS attack and
they can be mitigated by implementing some of the same best practices to be covered later in
this paper.

Application layer attacks can also target domain name system (DNS) services. The most
common of these attacks is a DNS query flood in which an attacker uses many well-formed
DNS queries to exhaust the resources of a DNS server. These attacks can also include a cache-
busting component where the attacker randomizes the subdomain string to bypass the local
DNS cache of any given resolver. As a result, the resolver can’t take advantage of cached
domain queries and must instead repeatedly contact the authoritative DNS server, which
amplifies the attack.

If a web application is delivered over TLS, an attacker can also choose to attack the TLS
negotiation process. TLS is computationally expensive so an attacker can reduce a server’s
availability by sending unintelligible data. In a variation of this attack, an attacker completes
the TLS handshake but perpetually renegotiates the encryption method. Or an attacker can
attempt to exhaust server resources by opening and closing many TLS sessions.

Page 6
Amazon Web Services – AWS Best Practices for DDoS Resiliency

Mitigation Techniques

Some forms of DDoS mitigation are included automatically with AWS services. You can further
improve your DDoS resilience by using an AWS architecture with specific services and by
implementing additional best practices.

All AWS customers benefit from the automatic protections of AWS Shield Standard, at no
additional charge. AWS Shield Standard defends against most common, frequently occurring
network and transport layer DDoS attacks that target your web site or applications. This is
offered on all AWS services and in every AWS Region, at no additional cost. In AWS Regions,
DDoS attacks are detected by a system that automatically baselines traffic, identifies
anomalies, and, as necessary, creates mitigations. This mitigation system provides protection
against many common infrastructure layer attacks. You can use AWS Shield Standard as part of
a DDoS-resilient architecture to protect both web and non-web applications.

Additionally, you can leverage AWS services that operate from edge locations, like Amazon
CloudFront and Amazon Route 53, to build comprehensive availability protection against all
known infrastructure layer attacks. Using these services – part of the AWS Global Edge
Network – can improve the DDoS resilience of your application when you serve web
application traffic from edge locations distributed around the world.

Several specific benefits of using Amazon CloudFront and Amazon Route 53 include the
following:

        AWS Shield DDoS mitigation systems that are integrated with AWS edge services,
         reducing time-to-mitigate from minutes to sub-second.

        Stateless SYN Flood mitigation techniques that proxy and verify incoming connections
         before passing them to the protected service.

        Automatic traffic engineering systems that can disperse or isolate the impact of large
         volumetric DDoS attacks.

        Application layer defense when combined with AWS WAF that does not require
         changing your current application architecture (e.g. in an AWS Region or on-premises
         datacenter).

Page 7
Amazon Web Services – AWS Best Practices for DDoS Resiliency

There is no charge for inbound data transfer on AWS and you do not pay for DDoS attack
traffic that is mitigated by AWS Shield.

For a DDoS-resilient reference architecture that includes AWS Global Edge Network services,
see the following diagram (Figure 5).

                           Figure 5: DDoS-resilient reference architecture.

This reference architecture includes several AWS services that can help you improve your web
application’s resiliency against DDoS attacks. For a summary of these services and the
capabilities that they can provide, see the following table (Table 2). We’ve tagged each service
with a “best practice indicator” (BP1, BP2, etc.) so that you can easily refer back to each of
them here as you read the rest of this document. For example, an upcoming section discusses
the capabilities provided by Amazon CloudFront and includes the best practices indicator BP1.

                       AWS Edge Locations                               AWS Regions

                       Amazon             Amazon      Elastic      Amazon     Amazon   Amazon EC2
                       CloudFront         Route 53    Load         API        VPC      with Auto
                       (BP1) with         (BP3)       Balancing    Gateway    (BP5)    Scaling (BP7)
                       AWS WAF                        (BP6)        (BP4)
                       (BP2)
  Layer 3 (e.g. UDP    ✔                  ✔           ✔            ✔          ✔        ✔
  reflection) attack
  mitigation

Page 8
Amazon Web Services – AWS Best Practices for DDoS Resiliency

                       AWS Edge Locations                           AWS Regions

  Layer 4 (e.g. SYN    ✔               ✔           ✔           ✔
  flood) attack
  mitigation

  Layer 6 (e.g. TLS)   ✔                           ✔           ✔
  attack mitigation

  Reduce attack        ✔               ✔           ✔           ✔        ✔
  surface

  Scale to absorb      ✔               ✔           ✔           ✔                  ✔
  application layer
  traffic

  Layer 7              ✔               ✔           ✔(if used
  (application                                     with AWS
                                                   WAF)
  layer) attack
  mitigation

  Geographic           ✔               ✔
  isolation and
  dispersion of
  excess traffic,
  and larger DDoS
  attacks

                              Table 2: Summary of best practices.

Another way that you can improve your readiness to respond to and mitigate DDoS attacks is
by subscribing to AWS Shield Advanced. This optional DDoS mitigation service helps you
protect an application hosted on any AWS Region or hosted outside of AWS. The service is
available globally for Amazon CloudFront and Amazon Route 53. It’s also available in select
AWS Regions for Classic Load Balancer (CLB), Application Load Balancer (ALB), and Elastic IP
Addresses (EIPs). Using AWS Shield Advanced with EIPs allows you to protect Network Load
Balancer (NLBs) or Amazon EC2 instances.

With AWS Shield Advanced, you get the following additional benefits:

        Access to the AWS DDoS Response Team (DRT) for assistance in mitigating DDoS
         attacks that impact application availability.

Page 9
Amazon Web Services – AWS Best Practices for DDoS Resiliency

         DDoS attack visibility by using the AWS Management Console, API, and Amazon
          CloudWatch metrics and alarms.

         Access to the Global Threat Environment dashboard, which provides an overview of
          DDoS attacks observed and mitigated by AWS.

         Access to AWS WAF, at no additional cost, for the mitigation of application layer DDoS
          attacks (when used with Amazon CloudFront or ALB).

         Automatic baselining of web traffic attributes, when used with AWS WAF.

         Access to AWS Firewall Manager, at no additional cost, for automated policy
          enforcement. This service lets security administrators centrally control and manage
          AWS WAF rules.

         Sensitive detection thresholds which routes traffic into DDoS mitigation system
          earlier and can improve time-to-mitigate attacks against Amazon EC2 or NLB, when
          used with EIP.

         Cost Protection that allows you to request a limited refund of scaling-related costs
          that result from a DDoS attack.

         Enhanced Service Level Agreement that is specific to AWS Shield Advanced
          customers.

For a complete list of AWS Shield Advanced features and to learn more about AWS Shield, see
AWS Shield – Managed DDoS Protection.

In the following sections, we’ll describe each of the recommended best practices for DDoS
mitigation in more depth. For a quick and easy-to-implement guide on building a DDoS
mitigation layer for static or dynamic web applications, see How to Help Protect Dynamic Web
Applications Against DDoS Attacks by Using Amazon CloudFront and Amazon Route 53.

Infrastructure Layer Defense (BP1, BP3, BP6, BP7)

In a traditional datacenter environment, you can mitigate infrastructure layer DDoS attacks by
using techniques like overprovisioning capacity, deploying DDoS mitigation systems, or
scrubbing traffic with the help of DDoS mitigation services. On AWS, DDoS mitigation
capabilities are automatically provided—but you can optimize your application’s DDoS-
resilience by making architecture choices that best leverage those capabilities and also allow
you to scale for excess traffic.

Page 10
Amazon Web Services – AWS Best Practices for DDoS Resiliency

Key considerations to help mitigate volumetric DDoS attacks include ensuring that enough
transit capacity and diversity is available, and protecting your AWS resources, like Amazon EC2
instances, against attack traffic.

Instance Size (BP7)
Amazon EC2 provides resizable compute capacity so that you can quickly scale up or down as
your requirements change. You can scale horizontally by adding instances to your application,
and scale vertically by using larger instances. Some Amazon EC2 instance types support
features that can more easily handle large volumes of traffic, for example, 25 Gigabit network
interfaces and Enhanced Networking.

With 25 Gigabit network interfaces, each instance can support a larger volume of traffic. This
helps prevent interface congestion for traffic that has reached the Amazon EC2 instance.
Instances that support Enhanced Networking provide higher I/O performance and lower CPU
utilization compared to traditional implementations. This improves the ability of the instance
to handle traffic with larger packet volumes.

The 25 Gigabit feature is available on the largest sizes of instances, for example, M4. To learn
more about Amazon EC2 instances that support 25 Gigabit network interfaces and Enhanced
Networking, see Amazon EC2 Instance Types. To learn how to enable Enhanced Networking,
see Enabling Enhanced Networking on Linux Instances in a VPC.

Choice of Region (BP7)
AWS services are available in multiple locations worldwide. These geographically-separate
areas are called regions. When architecting your application, you can choose one or more
regions based on your requirements. Common considerations include performance, cost, and
data sovereignty. In each region, AWS provides access to a unique set of internet connections
and peering relationships to provide optimal latency and throughput to users in those areas.

It is also important to consider DDoS resiliency when you choose the regions for your
application. Many regions are close to internet exchanges so they have more connectivity to
major networks. Being close to exchanges where international carriers and large peers have a
strong presence can help give you the internet capacity to mitigate larger attacks.

To learn more about choosing a region, see Regions and Availability Zones. You can also ask
your account team about the characteristics of each region that you’re considering, to help
you make an informed decision.

Page 11
Amazon Web Services – AWS Best Practices for DDoS Resiliency

Load Balancing (BP6)
Large DDoS attacks can overwhelm the capacity of a single Amazon EC2 instance, so adding
load balancing can help your resiliency. There are several options that you can choose from to
help mitigate an attack by load balancing excess traffic. With Elastic Load Balancing (ELB), you
can reduce the risk of overloading your application by distributing traffic across many backend
instances. ELB can scale automatically, allowing you to manage larger volumes when you have
unanticipated extra traffic, for example, due to flash crowds or DDoS attacks. For applications
built within Amazon VPC, there are two types of ELBs to consider, depending on your
application type: Application Load Balancer (ALB) or Network Load Balancer (NLB).

For web applications, you can use ALB to route traffic based on its content and accept only
well-formed web requests. This means that many common DDoS attacks, like SYN floods or
UDP reflection attacks, will be blocked by ALB, protecting your application from the attack.
When ALB detects these types of attacks, it automatically scales to absorb the additional
traffic. To learn more about protecting web applications with ALB, see Getting Started with
Application Load Balancers.

For TCP-based applications, you can use NLB to route traffic to Amazon EC2 instances at ultra-
low latency. When you create an NLB, a network interface is created for each Availability Zone
(AZ) that you enable. You have the option to assign an Elastic IP (EIP) address per subnet
enabled for the load balancer. One key consideration with NLB is that any traffic that reaches
the load balancer on a valid listener will be routed to your Amazon EC2 instances, not
absorbed. To learn more about protecting TCP applications with NLB, see Getting Started with
Network Load Balancers.

Deliver at Scale Using AWS Edge Locations (BP1, BP3)
Access to highly-scaled, diverse internet connections can significantly increase your ability to
optimize latency and throughput to users, to absorb DDoS attacks, and to isolate faults while
minimizing the impact on your application’s availability. AWS edge locations provide an
additional layer of network infrastructure that provides these benefits to any web application
that uses Amazon CloudFront and Amazon Route 53. When you use these services, your
content is served and DNS queries are resolved from locations that are often closer to your
users.

Web Application Delivery at the Edge (BP1)

Amazon CloudFront is a service that can be used to deliver your entire website, including
static, dynamic, streaming, and interactive content. Persistent TCP connections and variable
time-to-live (TTL) settings can be used to offload traffic from your origin, even if you are not

Page 12
Amazon Web Services – AWS Best Practices for DDoS Resiliency

serving cacheable content. These features mean that using Amazon CloudFront reduces the
number of requests and TCP connections back to your origin which helps protect your web
application from HTTP floods. Amazon CloudFront only accepts well-formed connections,
which helps prevent many common DDoS attacks, like SYN floods and UDP reflection attacks,
from reaching your origin. DDoS attacks are also geographically isolated close to the source
which prevents the traffic from impacting other locations. These capabilities can greatly
improve your ability to continue serving traffic to users during large DDoS attacks. You can use
Amazon CloudFront to protect an origin on AWS or elsewhere on the internet.

If you’re using Amazon S3 to serve static content on the internet, you should use Amazon
CloudFront to protect your bucket. You can use Origin Access Identify (OAI) to ensure that
users only access your objects by using CloudFront URLs. To learn more about OAI, see Using
an Origin Access Identity to Restrict Access to Your Amazon S3 Content.

To learn more about protecting and optimizing the performance of web applications using
Amazon CloudFront, see Getting Started with CloudFront.

Domain Name Resolution at the Edge (BP3)

Amazon Route 53 is a highly available and scalable domain name system (DNS) service that can
be used to direct traffic to your web application. It includes advanced features like Traffic Flow,
Latency Based Routing, Geo DNS, and Health Checks and Monitoring that allow you to control
how the service responds to DNS requests, to improve the performance of your web
application and to avoid site outages.

Amazon Route 53 uses techniques like shuffle sharding and anycast striping, that can help
users access your application even if the DNS service is targeted by a DDoS attack. With shuffle
sharding, each name server in your delegation set corresponds to a unique set of edge
locations and internet paths. This provides greater fault tolerance and minimizes overlap
between customers. If one name server in the delegation set is unavailable, users can retry
and receive a response from another name server at a different edge location. Anycast striping
allows each DNS request to be served by the most optimal location, spreading the network
load and reducing DNS latency. In turn, this provides a faster response for users. Additionally,
Amazon Route 53 can detect anomalies in the source and volume of DNS queries, and
prioritize requests from users that are known to be reliable.

To learn more about using Amazon Route 53 to route users to your application, see Getting
Started with Amazon Route 53.

Page 13
Amazon Web Services – AWS Best Practices for DDoS Resiliency

Application Layer Defense (BP1, BP2, BP6)

Many of the techniques discussed in this paper are effective at mitigating the impact that
infrastructure layer DDoS attacks have on your application’s availability. To also defend against
application layer attacks requires you to implement an architecture that allows you to
specifically detect, scale to absorb, and block malicious requests. This is an important
consideration because network-based DDoS mitigation systems are generally ineffective at
mitigating complex application layer attacks.

Detect and Filter Malicious Web Requests (BP1, BP2)
When your application runs on AWS, you can leverage both Amazon CloudFront and AWS WAF
to help defend against application layer DDoS attacks.

Amazon CloudFront allows you to cache static content and serve it from AWS edge locations,
which can help reduce the load on your origin. It can also help reduce server load by
preventing non-web traffic from reaching your origin. Additionally, CloudFront can
automatically close connections from slow-reading or slow-writing attackers (e.g. Slowloris).

By using AWS WAF, you can configure web access control lists (Web ACLs) on your CloudFront
distributions or Application Load Balancers to filter and block requests based on request
signatures. Each Web ACL consists of rules that you can configure to string match or regex
match one or more request attributes, such as the URI, query string, HTTP method, or header
key. In addition, by using AWS WAF's rate-based rules, you can automatically block the IP
addresses of bad actors when requests matching a rule exceed a threshold that you define.
Requests from offending client IP addresses will receive “403 Forbidden” error responses and
will remained blocked until request rates drop below the threshold. This is useful for
mitigating HTTP flood attacks that are disguised as regular web traffic.

To block attacks from known bad-acting IP addresses, you can create rules using IP match
conditions or use Managed Rules for AWS WAF offered by sellers in the AWS Marketplace that
will block specific malicious IP addresses that are included in IP reputation lists. Both AWS WAF
and Amazon CloudFront also allow you to set geo-restrictions to block or whitelist requests
from selected countries. This can help block attacks originating from geographic locations
where you do not expect to serve users.

To help identify malicious requests, you can review your web server logs or use WAF’s
Sampled Requests feature which provides details about requests that have matched one of
your AWS WAF rules for a period of time in the past 3 hours. You can use this information to
identify potentially malicious traffic signatures and create a new rule to deny those requests. If
you see a number of requests with a random query string, you might decide to disable query

Page 14
Amazon Web Services – AWS Best Practices for DDoS Resiliency

string forwarding in Amazon CloudFront. This can be helpful in mitigating a cache-busting
attack against your origin.

If you are subscribed to AWS Shield Advanced, you can engage the AWS DDoS Response Team
(DRT) to help you create rules to mitigate an attack that is hurting your application’s
availability. DRT can only gain limited access to your account, and only with your explicit
authorization. For more information, see the Support section later in this document.

Using AWS Firewall Manager can help simplify managing AWS WAF rules across your
organization. By using AWS Firewall Manager, you can enable AWS WAF across many accounts
and resources, including creating rules that are automatically applied to existing or new
accounts in your organization.

To learn more about using geo restriction to limit access to your Amazon CloudFront
distribution, see Restricting the Geographic Distribution of Your Content.

To learn more about using AWS WAF, see Getting Started with AWS WAF and Viewing a
Sample of the Web Requests that CloudFront has Forwarded to AWS WAF.

To learn more about configuring rate-based rules, see Protect Web Sites & Services Using
Rate-Based Rules for AWS WAF.

To learn how to manage the deployment of AWS WAF rules across your AWS resources with
AWS Firewall Manager, see Getting Started with AWS Firewall Manager.

Scale to Absorb (BP6)
Another way to mitigate application layer attacks is to operate at scale. If you have web
applications, you can use ELB to distribute traffic to a number of Amazon EC2 instances that
are overprovisioned or configured to automatically scale. These instances can handle sudden
traffic surges that occur for any reason, including a flash crowd or an application layer DDoS
attack. You can set Amazon CloudWatch alarms to initiate Auto Scaling to automatically scale
the size of your Amazon EC2 fleet in response to events that you define. This protects
application availability when there’s an unexpected increase in request volume. If you use
Amazon CloudFront or Application Load Balancer (ALB) with your application, TLS negotiation
is handled by the distribution or the load balancer. This helps protect your instances from
being impacted by TLS-based attacks by scaling to handle legitimate requests as well as TLS-
abuse attacks.

To learn more about using Amazon CloudWatch to invoke Auto Scaling, see Monitoring Your
Auto Scaling Instances and Groups Using Amazon CloudWatch.

Page 15
Amazon Web Services – AWS Best Practices for DDoS Resiliency

Attack Surface Reduction

Another important consideration when architecting an AWS solution is to limit the
opportunities an attacker has for targeting your application. For example, if you do not expect
users to directly interact with certain resources, you can make sure that those resources are
not accessible from the internet. Similarly, if you do not expect users or external applications
to communicate with your application on certain ports or protocols, you can make sure that
that traffic is not accepted.

This concept is known as attack surface reduction. In this section, we provide best practices to
help you reduce your attack surface and limit your application’s internet exposure. Resources
that are not exposed to the internet are more difficult to attack, which limits the options an
attacker has to target your application’s availability.

Obfuscating AWS Resources (BP1, BP4, BP5)

Typically, users can quickly and easily use an application without requiring that AWS resources
be fully exposed to the internet. For example, when you have Amazon EC2 instances behind an
ELB, the instances themselves might not need to be publicly accessible. Instead, you could
provide users with access to the ELB on certain TCP ports and allow only the ELB to
communicate with the instances. You can set this up by configuring Security Groups and
Network Access Control Lists (NACLs) within your Amazon Virtual Private Cloud (VPC). Amazon
VPC allows you to provision a logically isolated section of the AWS cloud where you can launch
AWS resources in a virtual network that you define.

Security groups and NACLs are similar in that they allow you to control access to AWS
resources within your VPC. But security groups allow you to control inbound and outbound
traffic at the instance level, while NACLs offer similar capabilities at the VPC subnet level.
There is no additional charge for using security groups or NACLs.

Security Groups and Network Access Control Lists (NACLs) (BP5)
You can specify security groups when you launch an instance, or you can associate the
instance with a security group at a later time. All internet traffic to a security group is implicitly
denied unless you create an allow rule to permit the traffic. For example, if you have a web
application that uses an ELB and a number of Amazon EC2 instances, you might decide to
create one security group for the ELB (“ELB security group”) and one for the instances (“web
application server security group”). You can then create an allow rule to permit internet traffic

Page 16
Amazon Web Services – AWS Best Practices for DDoS Resiliency

to the ELB security group, and another rule to permit traffic from the ELB security group to the
web application server security group. This ensures that internet traffic can’t directly
communicate with your Amazon EC2 instances, which makes it more difficult for an attacker to
learn about and impact your application.

When you create NACLs, you can specify both allow and deny rules. This is useful if you want
to explicitly deny certain types of traffic to your application. For example, you can define IP
addresses (as CIDR ranges), protocols, and destination ports that are denied access to the
entire subnet. If your application is used only for TCP traffic, you can create a rule to deny all
UDP traffic, or vice versa. This option is useful when responding to DDoS attacks because it lets
you create your own rules to mitigate the attack when you know the source IPs or other
signature.

If you are subscribed to AWS Shield Advanced, you can register Elastic IPs (EIPs) as Protected
Resources. DDoS attacks against EIPs that have been registered as Protected Resources are
detected more quickly, which can result in a faster time-to-mitigate. When an attack is
detected, the DDoS mitigation systems read the NACL that corresponds to the targeted EIP
and enforce it at the AWS network border. This significantly reduces your risk of impact from a
number of infrastructure layer DDoS attacks.

To learn more about configuring Security Groups and NACLs to optimize for DDoS resiliency,
see How to Help Prepare for DDoS Attacks by Reducing Your Attack Surface.

To learn more about using AWS Shield Advanced with EIPs as Protected Resources, see Enable
and Configure AWS Shield Advanced.

Protecting Your Origin (BP1, BP5)
If you’re using Amazon CloudFront with an origin that is inside of your VPC, you should use an
AWS Lambda function to automatically update your security group rules to allow only Amazon
CloudFront traffic. This improves your origin’s security by helping to ensure that malicious
users cannot bypass Amazon CloudFront and AWS WAF when accessing your web application.

To learn more about how to protect your origin by automatically updating your security
groups, see How to Automatically Update Your Security Groups for Amazon CloudFront and
AWS WAF by Using AWS Lambda.

You may also want to ensure that only your Amazon CloudFront distribution can forward
requests to your origin. With Edge-to-Origin Request Headers, you can add or override the
value of existing request headers when Amazon CloudFront forwards requests to your origin.
You can use the X-Shared-Secret header to help validate that requests made to your origin
were sent from Amazon CloudFront.

Page 17
Amazon Web Services – AWS Best Practices for DDoS Resiliency

To learn more about protecting your origin with an X-Shared-Secret header, see Forwarding
Custom Headers to Your Origin.

Protecting API Endpoints (BP4)
Typically, when you must expose an API to the public, there is a risk that the API frontend
could be targeted by a DDoS attack. To help reduce the risk, you can use Amazon API Gateway
as a “front door” to applications running on Amazon EC2, AWS Lambda, or elsewhere. By using
Amazon API Gateway, you don’t need your own servers for the API frontend and you can
obfuscate other components of your application. By making it harder to detect your
application’s components, you can help prevent those AWS resources from being targeted by
a DDoS attack.

When you use Amazon API Gateway, you can choose from two types of API endpoints. The
first is the default option: edge-optimized API endpoints that are accessed through an Amazon
CloudFront distribution. The distribution is created and managed by API Gateway, however, so
you don’t have control over it. The second option is to use a regional API endpoint that is
accessed from the same AWS region in which your REST API is deployed. We recommend that
you use the second type of endpoint, and then associate it with your own Amazon CloudFront
distribution. By doing this, you have control over the Amazon CloudFront distribution and the
ability to use AWS WAF for application layer protection.

When you use Amazon CloudFront and AWS WAF with Amazon API Gateway, configure the
following options:

      Configure the cache behavior for your distributions to forward all headers to the API
       Gateway regional endpoint. By doing this, CloudFront will treat the content as dynamic
       and skip caching the content.

      Protect your API Gateway against direct access by configuring the distribution to
       include the origin custom header x-api-key, by setting the API key value in API
       Gateway.

      Protect your backend from excess traffic by configuring standard or burst rate-limits
       for each method in your REST APIs.

       To learn more about creating APIs with Amazon API Gateway, see Getting Started with
       Amazon API Gateway.

Page 18
Amazon Web Services – AWS Best Practices for DDoS Resiliency

Operational Techniques

The mitigation techniques in this paper help you architect applications that are inherently
resilient against DDoS attacks. In many cases, it’s also useful to know when a DDoS attack is
targeting your application so you can take mitigation steps. If you experience an attack, you
may also benefit from help in assessing the threat and reviewing the architecture of your
application, or you might want to request other assistance. This section discusses best
practices for gaining visibility into abnormal behavior, alerting and automation, and engaging
AWS for additional support.

Visibility

When a key operational metric deviates substantially from the expected value, an attacker
may be attempting to target your application’s availability. If you’re familiar with the normal
behavior of your application, you can more quickly take action when you detect an anomaly.

By using Amazon CloudWatch, you can monitor applications that you run on AWS. For
example, you can collect and track metrics, collect and monitor log files, set alarms, and
automatically respond to changes in your AWS resources.

If you have architected your application by following the DDoS-resilient reference architecture,
common infrastructure layer attacks will be blocked before reaching your application. If you
are subscribed to AWS Shield Advanced, you have access to a number of CloudWatch metrics
that can indicate that your application is being targeted. For example, you can configure
alarms to notify you when there is a DDoS attack in progress, so you can check your
application’s health and decide whether to engage DRT. You can configure the DDoSDetected
metric to tell you if an attack has been detected. If you want to be alerted based on the attack
volume, you can also use the DDoSAttackBitsPerSecond, DDoSAttackPacketsPerSecond, or
DDoSAttackRequestsPerSecond metrics. You can monitor these metrics by integrating Amazon
CloudWatch with your own tools or by using tools provided by third-parties, such as Slack or
PagerDuty.

An application layer attack can elevate many Amazon CloudWatch metrics. If you’re using AWS
WAF, you can use CloudWatch to monitor and alarm on increases in requests that you’ve set in

Page 19
Amazon Web Services – AWS Best Practices for DDoS Resiliency

WAF to be allowed, counted, or blocked. This allows you to receive a notification if the level of
traffic exceeds what your application can handle. You can also use Amazon CloudFront,
Amazon Route 53, ALB, NLB, Amazon EC2, and Auto Scaling metrics that are tracked in
CloudWatch to detect changes that can indicate a DDoS attack.

For a description of Amazon CloudWatch metrics that are commonly used to detect and react
to DDoS attacks, see the following table (Table 3).

  Topic          Metric                           Description

  AWS Shield     DDoSDetected                     Indicates a DDoS event for a specific Amazon
  Advanced                                        Resource Name (ARN).

  AWS Shield     DDoSAttackBitsPerSecond          The number of bytes observed during a DDoS
  Advanced                                        event for a specific Amazon Resource Name
                                                  (ARN). This metric is only available for layer 3/4
                                                  DDoS events.

  AWS Shield     DDoSAttackPacketsPerSecond       The number of packets observed during a DDoS
  Advanced                                        event for a specific Amazon Resource Name
                                                  (ARN). This metric is only available for layer 3/4
                                                  DDoS events.

  AWS Shield     DDoSAttackRequestsPerSecond      The number of requests observed during a DDoS
  Advanced                                        event for a specific Amazon Resource Name
                                                  (ARN). This metric is only available for layer 7
                                                  DDoS events and is only reported for the most
                                                  significant layer 7 events.

  AWS WAF        AllowedRequests                  The number of allowed web requests.

  AWS WAF        BlockedRequests                  The number of blocked web requests.

  AWS WAF        CountedRequests                  The number of counted web requests.

  Amazon         Requests                         The number of HTTP/S requests
  CloudFront

  Amazon         TotalErrorRate                   The percentage of all requests for which the
  CloudFront                                      HTTP status code is 4xx or 5xx.

  Amazon         HealthCheckStatus                The status of the health check endpoint.
  Route 53

  ALB            ActiveConnectionCount            The total number of concurrent TCP connections
                                                  that are active from clients to the load balancer,
                                                  and from the load balancer to targets.

  ALB            ConsumedLCUs                     The number of load balancer capacity units (LCU)
                                                  used by your load balancer.

Page 20
Amazon Web Services – AWS Best Practices for DDoS Resiliency

  Topic         Metric                        Description

  ALB           HTTPCode_ELB_4XX_Count        The number of HTTP 4xx or 5xx client error codes
                HTTPCode_ELB_5XX_Count        generated by the load balancer.

  ALB           NewConnectionCount            The total number of new TCP connections
                                              established from clients to the load balancer, and
                                              from the load balancer to targets.

  ALB           ProcessedBytes                The total number of bytes processed by the load
                                              balancer.

  ALB           RejectedConnectionCount       The number of connections that were rejected
                                              because the load balancer had reached its
                                              maximum number of connections.

  ALB           RequestCount                  The number of requests that were processed.

  ALB           TargetConnectionErrorCount    The number of connections that were not
                                              successfully established between the load
                                              balancer and the target.

  ALB           TargetResponseTime            The time elapsed, in seconds, after the request
                                              left the load balancer until a response from the
                                              target was received.

  ALB           UnHealthyHostCount            The number of targets that are considered
                                              unhealthy.

  NLB           ActiveFlowCount               The total number of concurrent TCP flows (or
                                              connections) from clients to targets.

  NLB           ConsumedLCUs                  The number of load balancer capacity units (LCU)
                                              used by your load balancer.

  NLB           NewFlowCount                  The total number of new TCP flows (or
                                              connections) established from clients to targets in
                                              the time period.

  NLB           ProcessedBytes                The total number of bytes processed by the load
                                              balancer, including TCP/IP headers.

  Auto          GroupMaxSize                  The maximum size of the Auto Scaling group
  Scaling

  Amazon        CPUUtilization                The percentage of allocated EC2 compute units
  EC2                                         that are currently in use.

  Amazon        NetworkIn                     The number of bytes received by the instance on
  EC2                                         all network interfaces.

                     Table 3: Recommended Amazon CloudWatch metrics.

To learn more about using Amazon CloudWatch to detect DDoS attacks on your application,
see Getting Started with Amazon CloudWatch.

Page 21
Amazon Web Services – AWS Best Practices for DDoS Resiliency

AWS includes several additional metrics and alarms to notify you about an attack and to help
you monitor your application’s resources. The AWS Shield console or API provide a summary
and details about attacks that have been detected. In addition, the Global Threat Environment
Dashboard provides summary information about all DDoS attacks that have been detected by
AWS. This can be useful in better understanding DDoS threats across a larger population of
applications, understanding attack trends, and comparing with attacks that you may have
observed.

Another tool that can help you gain visibility into traffic that is targeting your application is VPC
Flow Logs. On a traditional network, you might use network flow logs to troubleshoot
connectivity and security issues, and to make sure that network access rules are working as
expected. By using VPC Flow Logs, you can capture information about the IP traffic that is
going to and from network interfaces in your VPC.

Each flow log record includes the following: source and destination IP addresses, source and
destination ports, protocol, and the number of packets and bytes transferred during the
capture window. You can use this information to help identify anomalies in network traffic and
to identify a specific attack vector. For example, most UDP reflection attacks have specific
source ports, such as source port 53 for DNS reflection. This is a clear signature that you can
identify in the flow log record. In response, you might choose to block the specific source port
at the instance level or create a NACL rule to block the entire protocol if your application
doesn’t require it.

To learn more about using VPC Flow Logs to identify network anomalies and DDoS attack
vectors, see VPC Flow Logs and VPC Flow Logs – Log and View Network Traffic Flows.

Support

It is important to create a response plan for DDoS attacks before an actual event. The best
practices outlined in this paper are intended to be proactive measures that you implement
before you launch an application, but DDoS attacks against your application might still occur.
Review the options in this section to determine the support resources that are best for your
scenario. Your account team can evaluate your use case and application, and assist with
specific questions or challenges that you have.

If you’re running production workloads on AWS, consider subscribing to Business Support
which provides you with 24 x 7 access to Cloud Support Engineers who can assist with DDoS
attack issues. If you’re running mission-critical workloads, consider Enterprise Support which
provides you with the ability to open “Critical” cases and receive the fastest response from a

Page 22
Amazon Web Services – AWS Best Practices for DDoS Resiliency

Senior Cloud Support Engineer.

If you’re subscribed to AWS Shield Advanced and are also subscribed to either Business
Support or Enterprise Support, you can escalate to the AWS DDoS Response Team (DRT) if you
have a DDoS-related event that impacts your application’s availability. If your application’s
responsiveness is degraded because of a DDoS attack, you can make live contact with AWS
Support. Another option is to use the AWS Shield Engagement Lambda function to more
quickly initiate contact with DRT. For example, you can use an AWS IoT button to trigger the
AWS Lambda function if you have an emergency situation. When you press the button, a case
is automatically opened with AWS Support and DRT is notified immediately. You receive a
direct reply for your case that includes an Amazon Chime conference bridge that you can join
to interact with AWS Support and DRT. The AWS Shield Engagement Lambda can be used with
any trigger supported by AWS Lambda.

To learn more about rapid DRT engagement by using an AWS Lambda function, see AWS Shield
Engagement Lambda.

DRT does not typically have access to your AWS account or AWS WAF Sampled Requests. You
can authorize DRT to access AWS WAF, AWS Shield, and related API operations on your
account from the AWS Shield console or API. For example, you may want to allow DRT to view
your Sampled Requests or place rules to assist with the mitigation of an application layer DDoS
attack. You can also authorize DRT to access Amazon S3 buckets that you specify. For example,
you may have a bucket where you store web request logs and would like DRT to have access to
them for analysis during an attack. DRT will only access your account or make changes during
an escalated event and any changes will be subject to your consent. To learn more about
granting limited account access to DRT, see Authorize the DDoS Response Team.

In some cases, DRT may learn about a DDoS attack and engage you proactively. If there are
specific points of contact who should be engaged during DRT-driven escalation, you can add
them in the AWS Shield console by clicking “Summary,” followed by “Edit” under the
“Additional contacts” section.

Page 23
Amazon Web Services – AWS Best Practices for DDoS Resiliency

Conclusion

The best practices outlined in this paper can help you to build a DDoS-resilient architecture
that can protect your application’s availability by preventing many common infrastructure and
application layer DDoS attacks. The extent to which you follow these best practices when you
architect your application will influence the type, vector, and volume of DDoS attacks that you
can mitigate. You can build in resiliency without subscribing to a DDoS mitigation service. Or,
optionally, you can choose to subscribe to AWS Shield Advanced to gain additional support,
visibility, mitigation, and cost protection features that further protect an already resilient
application architecture.

To learn more about DDoS mitigation and DDoS resiliency best practices on AWS, see
Appendix A: Additional Resources.

Contributors

The following individuals and organizations contributed to this document:

         Andrew Kiggins, AWS Solutions Architect

         Jeffrey Lyon, AWS Perimeter Protection

         Achraf Souk, AWS Solutions Architect

         Tino Tran, AWS Solutions Architect

         Yoshihisa Nakatani, AWS Solutions Architect

Document Revisions

 Date                Description

 June 2018           Updated to include AWS Shield, AWS WAF features, AWS Firewall Manager, and
                     related best practices.

 June 2016           Added prescriptive architecture guidance and updated to include AWS WAF.

Page 24
Amazon Web Services – AWS Best Practices for DDoS Resiliency

 Date            Description

 June 2015       First publication.

Page 25
Amazon Web Services – AWS Best Practices for DDoS Resiliency

Appendix A: Additional Resources
You can use the following resources to learn more about DDoS mitigation and DDoS-resiliency
best practices with AWS:

         Best Practices for DDoS Mitigation on AWS

         SID216 – re:Invent 2017: Cloud-Native App Protection: Web Application Security at
          Pearson

         SID324 – re:Invent 2017: Automating DDoS Response in the Cloud

         CTD304 – re:Invent 2017: Dow Jones & Wall Street Journal’s Journey to Manage
          Traffic Spikes While Mitigating DDoS & Application Layer Threats

         CTD310 – re:Invent 2017: Living on the Edge, It’s Safer Than You Think! Building
          Strong with Amazon CloudFront, AWS Shield, and AWS WAF

Page 26
You can also read